U.S. patent number 6,951,759 [Application Number 10/219,939] was granted by the patent office on 2005-10-04 for detection of bacterial vaginosis.
This patent grant is currently assigned to Osmetech PLC. Invention is credited to Amjad Nissar Chaudry, Martin James Henery, Paul James Travers, Andrew John Tummon.
United States Patent |
6,951,759 |
Travers , et al. |
October 4, 2005 |
Detection of bacterial vaginosis
Abstract
There is disclosed a method for detecting the presence of
bacterial vaginosis in a female subject comprising the steps of:
obtaining a sample from the vaginal region of the subject;
detecting acetic acid present in the sample; and correlating the
presence of detected acetic acid with the presence of bacterial
vaginosis.
Inventors: |
Travers; Paul James (Chorlton,
GB), Chaudry; Amjad Nissar (Chorlton, GB),
Tummon; Andrew John (Middlewich, GB), Henery; Martin
James (Alderley Edge, GB) |
Assignee: |
Osmetech PLC (Crewe,
GB)
|
Family
ID: |
9920563 |
Appl.
No.: |
10/219,939 |
Filed: |
August 14, 2002 |
Foreign Application Priority Data
|
|
|
|
|
Aug 17, 2001 [GB] |
|
|
0120062 |
|
Current U.S.
Class: |
436/129;
435/252.1; 435/807; 436/111; 436/181; 436/811; 600/562 |
Current CPC
Class: |
G01N
33/487 (20130101); Y10S 436/811 (20130101); Y10S
435/807 (20130101); Y10T 436/201666 (20150115); Y10T
436/173845 (20150115); Y10T 436/25875 (20150115) |
Current International
Class: |
G01N
33/52 (20060101); G01N 33/00 (20060101); G01N
033/00 () |
Field of
Search: |
;436/129,111,181,811
;600/562 ;435/252.1,807 |
References Cited
[Referenced By]
U.S. Patent Documents
Foreign Patent Documents
|
|
|
|
|
|
|
63255653 |
|
Oct 1988 |
|
JP |
|
01074442 |
|
Mar 1989 |
|
JP |
|
WO 01/13087 |
|
Feb 2001 |
|
WO |
|
Other References
Brand et al., "Trimethylamine The Substance Mainly Responsible For
the Fishy Odor Often Associated with Bacterial Vaginosis,"
Obstetrics & Gynecology, vol. 68, No. 5, pp. 682-685 (1986)
(abstract). .
Hillier, Sharon L., "Diagnostic microbiology of bacterial
vaginosis," Am. J. Obstet Gynecol, Aug. 1993, pp. 455-459. .
Spiegel, Carol A., et al., "Anaerobic Bacteria in Nonspecific
Vaginitis," The New England Journal of Medicine, vol. 303, No. 11,
Sep. 11, 1980, pp 601-606. .
Thomason, Jessica L, et al., "Is analysis of vaginal secretions for
volatile organic acids to detect bacterial vaginosis of any
diagnostic value?," Am. J. Obstet Gynecol, Dec. 1988, pp.
1509-1511. .
Stanck, Ronald et al., "High Performance Ion Exclusion
Chromatographic Characterization of the Vaginal Organic Acids in
Women with Bacterial Vaginosis," Biomedical Chromatography, vol. 6,
1992, pp. 231-235. .
Piot, P., et al., "The Vaginal Microbial Flora in the Non-Specific
Vaginitis," Eur. J. Clin. Microbiol, vol. 1, No. 5, Oct. 1982, p.
301-306. .
Ison, C.A., et al., "Non-volatile fatty acids in the diagnosis of
non-specific vaginitis," J. Clin. Pathol., vol. 36, 1983, pp.
1367-1370. .
Spiegel, Carol. A. et al, Mobiluncus gen. nov., Mobiluncus curtisii
subsp. curtisii sp. nov., Mobiluncus curtisii subsp. holmesii
subsp. nov., and Mobiluncus mulieris sp. no., Curved Rods from the
Human Vagina, International Journal of Systematic Bacteriology,
vol. 34, No. 2, Apr. 1984, pp. 177-184. .
Crawley, B. A., et al., "C-82. Microbial Growth, Antibiotic
Sensitivity Patterns and the Electronic Nose," Abstracts of the
General Meeting of the American Society For Microbiology, vol. 99,
199, p. 121. .
Chandiok, S., et al., "Screening for bacterial vaginosis: a novel
application of artificial nose technology," J. Clin. Pathol., vol.
50, 1997, pp. 790-791. .
Xu, Clarke X., et al., "Development of a diamine biosensor,"
Talanta, vol. 44, Jan. 1997, pp. 1625-1632. .
Chandiok, S.; Crawley, B.A.; Oppenheim, B.A.; Chadwick, P.R.;
Higgins, S. and Persaud, K.C., "Screening for bacterial vaginosis:
a novel application of artificial nose technology," Short Reports,
J. Clin Pathol 1997; 50:790-791..
|
Primary Examiner: Cole; Monique T.
Attorney, Agent or Firm: Marger Johnson & McCollom,
PC
Parent Case Text
RELATED APPLICATION DATA
This application claims priority from Great Britain Application No.
0120062.5, filed 17 Aug. 2001 and U.S. Provisional Application Ser.
No. 60/335,185, filed 15 Nov. 2001.
Claims
What is claimed is:
1. A method for detecting the presence of bacterial vaginosis in a
female subject comprising the steps of: obtaining a sample from the
vaginal region of the subject; detecting acetic acid present in the
sample; and correlating the presence of detected acetic acid with
the presence of bacterial vaginosis.
2. A method according to claim 1 in which acetic acid present as a
gas in a headspace associated with the sample is detected using a
detector which is sensitive to the presence of acetic acid.
3. A method according to claim 1 in which the sample is a swab
sample.
4. A method according to claim 3 in which a high vaginal swab
sample is obtained.
5. A method according to claim 1 in which ammonia and, optionally,
amine species present in the sample are detected, and the presence
of detected acetic acid, ammonia and, optionally, amine species is
correlated with the presence of bacterial vaginosis.
6. A method according to claim 1 in which ammonia and, optionally,
amine species present as gas in the headspace associated with the
sample are detected by the detector.
7. A method according to claim 2 in which the sample is a swab
sample.
8. A method according to claim 3 in which a gas sensitive
layer.
9. A method according to claim 2 in which the detector comprises
semiconducting organic polymer.
10. A method according to claim 2 in which the detector comprises
at least one conductimetric gas sensor having a gas sensitive layer
onto which gases adsorb and desorp, and in which analytes are
detected by: exposing the gas sensor to the headspace, thereby
allowing the adsorption of analytes present in the headspace onto
the gas sensitive layer; and making conductimetric measurements of
the sensor during a desorption phase in which there is nett
desorption of analyte from the gas sensitive layer.
11. A method according to claim 10 in which the conductimetric gas
sensor or sensors comprise semiconducting organic polymer.
12. A method according to claim 7 in which: data reduction of one
or more measurements of one or more calibration samples is
performed to provide reference scores and reference loadings; and
data reduction of the output of the detector performed using the
reference loadings.
13. A method according to claim 12 in which the data reduction
comprises a principal components analysis.
14. A method according to claim 12 in which the reference scores
and reference loadings relate to a co-ordinate system having an
axis which is correlated to the presence of acetic acid.
15. A method according to claim 14 in which the reference scores
and reference loadings relate to a co-ordinate system having one
axis which is correlated to the presence of acetic acid and another
axis which is correlated to the presence of ammonia and,
optionally, amine species.
16. A method according to claim 12 in which: the reference scores
and reference loadings relate to a co-ordinate system having a
species characteristic axis which is correlated to the presence of
a species; and data reduction of intensity data from the detector
is performed using the reference loadings so that position along
the species characteristic axis is related to the concentration in
the headspace of the species which is correlated to the species
characteristic axis.
Description
BACKGROUND OF THE INVENTION
This invention relates to the detection of bacterial vaginosis
using a gas detector.
Bacterial vaginosis (EV) is a well known, but not well understood
or well defined, condition which exhibits uncertain symptoms.
Numerous reports cite as much as 50% of the affected population
being asymptomatic. The remaining 50% of the population either go
undetected or present during routine examination for an associated
or uncorrelated problem.
Originally thought to be a benign infection recent studies have
linked the problem to increased risk of intra-amniotic infection,
choroamionitis, post-caesarean and post-partum endometritis,
adverse pregnancy outcome, pre-term labour and birth, premature
rupture of membranes at term and post-hysterectomy cuff
cellulitis.
BV is commonly thought to arise as a result of fluctuation of the
normal vaginal flora, In some cases the flora can fluctuate
naturally over the menstrual cycle with no adverse effects. It is
thought that one of the primary controlling mechanisms controlling
BV causative bacteria is the presence of adequate colonies of
Lactobacillus sp. that produce hydrogen peroxide. The most common
organisms associated with BV are: Gardnerella vaginalis,
Bacteroides (Prevotella) spp., Mobiuncus spp. and Mycoplasma
hominis. However, the presence or absence of these flora is not
reliably diagnostic.
Treatment after a correct diagnosis is usually quite effective and
usually comprises of treatment with oral doses of metronidazole.
Topical treatments with metronidazole or clindamycin are also
common. However, in Doctors' surgeries it is not uncommon for
general practitioners (GPs) to wrongly diagnose vaginal infections,
e.g., some cases of Candida are diagnosed as BV and vice versa. A
distinction between a yeast and bacterial infection is important,
as non-specific antibiotics can cause more problems for the
sufferer. For symptomatic patients, BV has an incidence of 40%
compared with Candida and Trichomonas. A swift, preferably in-situ,
diagnosis would enable immediate correct therapy (usually
comprising antibiotics in the case of DV) to be administered. In
GPs' surgeres, a means of differentiating a bacterial infection
from a yeast infection would enable the correct type of treatment
to be prescribed.
In fact, the consequences of BV are wide and varied and are not
completely understood. This is perhaps unsurprising given the
difficulties in getting reliable BV data for a population. The
primary challenge facing any prospective diagnostic technique (or
aid to diagnosis) is finding a unique indicator against which BV
may be detected. Currently the Amsel test is the benchmark for
determining the problem. The criteria for the test rely on at least
three out of four conditions being met. These are; pH of vaginal
fluid >4.5 Typical thin, homogenous vaginal discharge Release of
strong fishy smell on adding alkali (10% KOH) to a sample of
vaginal fluid (whiff test), Clue cells present on microscopic
examination of a wet mount of vaginal fluid.
It should be noted that the presence of trimethylamine (TMA) in
some BV samples is undisputed, and TMA is often cited as being the
volatile associated with the unpleasant fishy odour referred to
above.
Individually none of these tests are diagnostic. pH variation of
the vaginal fluid is nearly always present in BV positive patients
but it is a non-specific. test and the variation is equally likely
to be caused by another infection or problem. Additionally,
contamination of the sample by cervical mucus (typical pH 7) can
lead to false diagnoses in some cases. pH variation also occurs as
part of the natural menstrual cycle. Ethnic background is also a
factor affecting vaginal pH and his has been postulated as a reason
for the relatively higher number of black American women who
present with the disease. According to Hay, pH is highly sensitive
(97%) but very non-specific giving false positives in 47% of cases.
Conversely, discharge is very accurately recognised by clinicians
giving false positives at 3% but only has a specificity of 67%.
Following this, the "whiff" test also gives low false positives
(1%) but is non-specific (43%). Finally clue cells are typically
found in 81% of positive BV cases whilst 6% of non-BV cases have
positive clue cell tests. Other trials report variation on these
figures but all concur with the non-specificity and reliability of
any one individual test.
It is known from the applicant's International Application No. WO
95/33848 that microorganisms can be detected using arrays of gas
sensors to detect characteristic gases or vapours produced by the
microorganisms. An example of such an array is an array of
semiconducting organic polymer gas sensors. The applicant's
International Applications Nos. WO 98/29563 and WO 98/39470
describe further aspects and refinements to the technique, and
related developments. In general, the approach with arrays of gas
sensors is to utilise a large number (twenty, thirty or more) of
different gas sensors which possesses different but overlapping
sensitivities towards different gaseous species (so-called
"electronic noses"). Oases are recognised from the characteristic
"fingerprint" or pattern of response across the array. However,
detection can be difficult in a complex system having mixed
populations of microflora and microfaunae and/or systems in which
many volatile species are present.
Of the particular relevance to the present application are
International Application number WO 94/04916 and Chandiok et al (S.
Chandiok, B A Crawley, B A Oppenhein, P R Chadwick, S Higgins, and
K C Persaud, Journal of Clinical Pathology, 50 (1997) 790). Both
documents describe attempts to detect BV using arrays of
semiconducting organic polymers, and in both cases it is believed
that the gaseous species detected were ammonia and/or TMA. In fact,
WO 94/04916 specifically describes a process rather similar to the
whiff test discussed above, in which KOH is added to a sample,
thereby releasing volatile alkaline species into the gas phase. It
should be noted, when considering the present invention as
described below, that the sensors used in the investigation of
Chandiok et al are not sensitive to fatty acids such as acetic
acid.
SUMMARY OF THE INVENTION
The present invention overcomes the above-mentioned problems and
difficulties, and provides a quick, reliable and practical
screening technique for the detection of bacterial vaginosis. The
technique is easily automated and may be performed by unskilled
operatives with a minimum of technical back-up.
For the avoidance of doubt, the terms "gas" and "gases" are
understood to embrace all species present in the gas phase,
including volatile species emanating from liquids and sublimed
species emanating from solids.
According to a first aspect of the invention there is provided a
method for detecting the presence of bacterial vaginosis in a
female subject comprising the steps of:
obtaining a sample from the vaginal region of the subject;
detecting acetic acid present in the sample; and
correlating the presence of the detected acetic acid with the
presence of bacterial vaginosis.
Surprisingly, acetic acid has been found to be a "marker" species
indicative of BV, and has enabled the provision of a detection
technique having the aforesaid advantages.
In a preferred embodiment, acetic acid present as a gas in a
headspace associated with the sample is detected using a detector
which is sensitive to the presence of acetic acid.
The sample may be a swab sample, and a high vaginal swab sample
maybe obtained. A high vaginal swab is a swab taken from the
Posterior Formix, any of the three vaulted spaces at the top of the
vagina, or from around the cervix of the uterus. Thus, the sample
can be "indirect", in the sense that the sample comprises an object
which is physically removed from the vaginal region of the subject
to a sampling portion of the detector. Alternatively, it may be
possible to sample gases directly from the vaginal region of the
subject using the detector: this alternative is also within the
scope of the invention.
Ammonia and, optionally, amine species (such as TMA) present in the
sample maybe detected, and the presence of acetic acid, ammonia
and, optionally, amine species maybe correlated with the presence
of bacterial vaginosis. Preferably, ammonia and, optionally, amine
species present as gas in the headspace associated with the sample
are detected by the detector.
The detector may comprise semiconducting organic polymer.
The detector may comprise an array of gas sensors. An array, in the
context of the present invention, is two or more gas sensors. In
contrast to conventional electronic noses, it has been found that
arrays having only a small number of physically different gas
sensors can be used advantageously, For example, an array may
comprise five or fewer sensor types which are sensitive to acetic
acid, and five or fewer sensor types which are sensitive to ammonia
(and/or amines). As few as four different sensor types have been
found to be sufficient. It may be possible to use a single acetic
acid sensitive gas sensor in place of an array.
The array may comprise gas sensors having semiconducting organic
polymer as a gas sensitive layer.
The detector may comprise at least one conductimetnic gas sensor
having a gas sensitive layer onto which gases absorb and desorb,
and in which analytes are detected by:
exposing the gas sensor to the headspace, thereby allowing the
adsorption of analytes present in the headspace onto the gas
sensitive layer; and
making conductimetric measurements of the sensor during a
desorption phase in which there is nett desorption of analyte from
the gas sensitive layer. This approach has been found to be
extremely advantageous in terms of improving sensitivity and
reproducibility. The principal reason for this is that the effect
of water vapour appears to be substantially eliminated in the
desorption phase. This is a considerable advantage, and
significantly enhances the detection of fatty acids, ammonia and
amines. However, it will be appreciated that this approach is not
limited to the detection of these species, and nor is it limited to
the method of the present invention. Rather, the approach of making
conductimetric measurements in the desorption phase can be employed
as a general technique for detecting analytes using conductimetric
gas sensors which have a gas sensitive layer. Typically, a pulse of
gas from the headspace is flowed over the sensor, and the
desorption phase commences once this pulse of gas has finished
flowing over the sensor. The technique of making conductimetric
measurements in the desorption phase works best when analytes of
interest have significantly longer desorption times than water.
The conductimetric gas sensor or sensors may comprise
semiconducting organic polymer.
This method may further comprise the steps of:
performing a data reduction of one or more measurements of one or
more calibration samples to provide reference scores and reference
loadings; and
performing data reduction of the output of the detector using the
reference loadings.
In a preferred embodiment the data reduction comprises a principal
components analysis (PCA). Other forms of data reduction, such as
Sammon mapping, maybe feasible.
This approach permits very convenient and quick assessment of
whether a sample is infected on the basis of simple assessment
criteria. Furthermore, it is possible to calibrate the system and
to perform self-test protocols using its approach.
The reference scores and reference loadings may relate to a
co-ordinate system having an axis which is correlated to the
presence of acetic acid. The reference scores and reference
loadings may relate to a co-ordinate system having one axis which
is correlated to the presence of acetic acid and another axis which
is, correlated to the presence of ammonia and, optionally, amine
species.
The method may be one in which:
the reference scores and reference loadings relate to a co-ordinate
system having a species characteristic axis which is correlated to
the presence of a species; and
data reduction of intensity data from the detector is performed
using the reference loadings so that position along the species
characteristic axis is related to the concentration in the
headspace of the species which is correlated to the species
characteristic axis. In this way the intensity data can be related
to the concentration of marker species which in turn can be related
to the number of infecting organisms. In particular, it is possible
to observe if a threshold concentration has been crossed, allowing
the correlation of detector output with the presence of infection
to be made. It should be noted that in the prior art, intensity
data from electronic noses comprising arrays of gas sensors are
usually removed by normalisation before analysis such as PCA, so
that concentration independent "fingerprints" can he obtained.
According to a second aspect of the invention there is provided a
gas sensing system adapted to detect the presence of bacterial
vaginosis in a female subject by the method of the first aspect of
the invention comprising:
a detector which is sensitive to the presence of gaseous acetic
acid and adapted to sample a headspace associated with the sample;
and
analysis means for analysing the output of the detector and
correlating the presence of detected acetic acid with the presence
of bacterial vaginosis.
The analysis means may correlate the presence of acetic acid,
ammonia and, optionally, amine species with the presence of
bacterial vaginosis.
Methods and apparatus in accordance with the invention will now be
described with reference to the accompanying drawings, in
which:
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 shows (a) a schematic diagram of apparatus suitable for
identifying the presence of BV and (b) Sampling of a headspace:
FIG. 2 shows typical sensor responses a) to acetic acid of one
polymer type and b) to ammonia of another polymer type as a
function of time;
FIG. 3 shows a PCA transformation;
FIG. 4 shows a PCA calibration map;
FIG. 5 shows the calculation of projected PCA scores;
FIG. 6 is a graphical representation of ellipses projected onto the
PCA calibration map; and
FIG. 7 is a graphical representation of the resolution of sample
standard deviation along the PC1-PC2 axes.
DETAILED DESCRIPTION
FIG. 1(a) schematically depicts apparatus shown generally at 10,
for use with the method of the present invention. The apparatus
comprises a sample carousel 12 in which a number of sample vials
can be mounted and maintained at a constant temperature, for
example 40.degree. C. For simplicity, a single sample vial 14 is
shown in FIG. 1(a). The vial 14 contains a sample 16 such as a swab
taken from the vagina of a subject. Above the sample 16 is a
gaseous headspace 18 which contains inter alia volatile species
emanating from the sample 16.
The vial 14 has a septum 14a thereon which is pierced by a needle
20, the insertion of needle 20 into the vial 14 being performed
automatically by the apparatus 10. The needle 20 is shown in more
detail in FIG. 1(b). The needle 20 is of co-axial design, which
permits a carrier gas (such as air, nitrogen or a noble gas) to be
introduced to vial 14 via the inner lumen 20a of the needle 20.
Gases in the headspace 18 are entrained in tile flow of carrier
gas, which exits the vial 14 via the outer lumen 20b of the needle
20, and thereafter is flowed across a gas sensor array 22. In this
way, the headspace 18 is sampled by a gas detector 22, which in
this embodiment is a gas sensor array. It will be appreciated by
the skilled reader that there are many other ways in which the
headspace might be coupled to a gas detector. It may be possible to
directly sample gases from the vaginal region of a subject using a
suitable arrangement, such as a sampling probe coupled to the gas
detector. Additionally, the use of devices such as filters and
preconcentrators is feasible.
The gas sensor array 22 is selected so that it can detect acetic
acid. This species can been found to constitute a "marker" species
which can be indicative of BV infection. Other species can be
detected in order to augment the identification. Ammonia and amine
species such as TMA can be detected for this purpose. It is
possible that further "marker" species might be detected in
addition to acetic acid.
The output of the gas sensor array 22 is monitored and analysed by
control means 24 which comprise computer means or other
microprocessor-based analysis means. The control means 24 can also
control the operation of the carousel 12, the flow of carrier gas,
washing and calibration procedures, and the manner in which the gas
sensor array 22 is operated or interrogated. However, it is quite
possible to transfer data from the control means 24 to, for
example, a remote computer for analysis. In any event, some form of
analysis means is provided which is adapted to correlate the
presence of the detected acetic acid with the presence of the
infection. In this way, the sample 14 is screened for BV
infection.
The method of the present invention has been used to screen samples
for BV infection. The human vagina is a host to many species of
microorganisms, and the headspace associated with a vaginal sample
is itself complex. It is known from WO 95/33848 that microorganisms
can produce volatile species which are characteristic of the
microorganisms. What is not known from WO 95/33848 is how, with a
highly complex headspace associated with a vaginal sample, one can
identify the presence of BV infection from gases emanating from the
sample. Furthermore, WO 94/04916 and Chandiok et al. are concerned
with the detection of alkaline species such as ammonia and amines.
The present invention overturns this conventional wisdom.
The pH of the sample can be lowered, by the addition of an acid in
order to release acetic acid into the gaseous phase. However, it
has been found that typically quite large concentrations of acetic
acid are associated with positive BV samples (often at
concentrations of 500 ppm or greater) and therefore acidification
is not an essential element of the invention.
Optionally, a salt such as Na.sub.2 SO.sub.4 can be added in order
to displace less soluable volatiles, in particular organic species,
from solution and into the gaseous phase. Backwashing between
samples is advisable to prevent cross-contamination.
In a preferred embodiment, the gas detector is an array of gas
sensors, and in a particularly preferred embodiment the gas sensors
comprise semiconducting organic polymer gas sensors. However, in
principle, other forms of gas detector might be employed, provided
that they are sensitive to the marker species described above. Gas
detection techniques which are candidates for use in the present
invention include gas chromatography, mass spectrometry and
spectroscopic techniques such as IR spectroscopy. Other forms of
gas sensor array might be contemplated, such as arrays of metal
oxide sensors, SAW sensors, quartz resonators, "composite" sensors
of the type described generally in U.S. Pat. No. 5,571,401, and
arrays comprising mixtures thereof.
Embodiments of a preferred--but non-limiting-kind of gas detector
will now be described, namely arrays of semiconducting organic
polymer gas sensors. As discussed above, the traditional approach
with such arrays is to employ a large number (typically twenty or
more) of different sensors having different polymers and/or
different dopant counterions, thus producing an array in which the
individual gas sensors exhibit broad and overlapping sensitivities
towards a range of different gases. The same principle applies to
other arrays of gas sensors, such as metal oxide sensors. Devices
of this type are commonly referred to as "electronic noses".
In direct contrast, it has been found that screening for infection
according to the present invention can be advantageously performed
using an array which comprises a limited number of sensor types,
ie. sensors with physically different polymer/counterion
combinations. In one example, four types of selective conducting
polymer sensors have been developed and incorporated into a device.
Two of these are acid sensitive, one sensitive to ammonia, and the
other sensitive to ammonia and trimethyl amine. These four sensor
types are incorporated into a 48 sensor array, comprising 12 sets
of replicate sensors. The provision of 12 replicates of each sensor
type permits signal averaging over a large number of sensors.
Additionally, sets of replicate sensors allows the array to
function in the event that one or even more than one sensor in any
given replicate set malfunctions. Fewer sensors still might be
utilised, particularly if acetic acid alone is detected as a
"marker".
The changes in resistance of each sensor type in response to a
volatile sample are recorded with time, and are averaged for each
sensor type over the array. It has been observed that it is
possible to eliminate the effect of water vapour on the response of
the sensors by choosing a portion of the trace corresponding to the
desorption phase of the experiment. With acetic acid as the
analyte, it has been observed that there is undershoot in the
signal below the previous baseline (see FIG. 2a). This effect is
reproducible is a function of concentration of acetic acid, and is
a parameter due to the type of materials incorporated into the
sensor. The time course is primarily dependent on the sensor
kinetics, but carrier flow and header geometry will also have an
effect.
FIG. 2a shows a number of response profiles to acetic acid for a
semiconducting organic polymer sensor against time. The baseline
response is indicated at "A". During the period of time indicated
as "B", the sensors are exposed to a pulse of gas comprising acetic
acid entrained in a carrier gas. This can be regarded as an
"adsorption phase" during which there is nett adsorption of acetic
acid--and water--onto the sensors. After the gas pulse has
finished, there is a desorption phase, or recovery phase, which is
indicated as "C", during which there is a nett desorption of
analyte from the sensors. It can be seen that the response becomes
negative with respect to the baseline during the recovery phase.
With fatty acid analytes, signal averaged over the period C is a
function of the concentration of acid present in the headspace. The
responses shown at "D" and "E" relate to (standard) wash and
reference cycles, respectively.
It should be noted that generally similar responses are obtained
when the sensors are exposed to ammonia (FIG. 2b), ie. there are
distinct baseline, adsorption and recovery phases. However, the
response in the recovery phase in this instance remains positive
with respect to the baseline. Measurements made during the recovery
phase are also substantially free from interferences from
moisture.
Interference from moisture is a major limitation for a number of
gas sensing technologies which interrogate a gas sensitive layer of
some kind upon which analytes--and water vapour--can reversibly
adsorb. Semiconducting organic polymers are an example of such a
gas sensitive layer. The above described technique for rejecting
interfering signals due to moisture is of broad significance--not
only is the technique applicable in the context of screening for BV
infection, it can be utilised more widely in the detection of
analytes per se.
It is believed that the displacement of the sensor response from
the baseline during the recovery phase is a result of the analyte
still being bound at the polymer surface. As a result of
interactions between the bound analyte and the electronic structure
of the polymer, the polymer can be more doped (producing a negative
response) or less doped (producing a positive response) than when
the baseline measurements were made. It is believed that water
desorbs from the polymer surface very rapidly during the recovery
phase, and so most of the recovery phase is substantially moisture
free. However, these mechanisms are speculative in nature, and
should not be regarded as a limiting one.
It should be noted that analysis of the sensor response is not
restricted to the recovery phase. In an alternative configuration
the resistances recorded when the sensors are exposed to the
headspace from the sample (ie, the sample phase) could be compared
to the resistances recorded for an equivalent exposure to a
threshold standard. By determining whether he sample resistance
lies above or below the resistance values for the threshold
standard the BV diagnosis can be made. In this configuration the
calibration requirements for the system may be greatly reduced.
It should be noted that, whilst prior art semiconducting organic
polymer gas sensors generally show good sensitivity towards
ammonia, it has not previously been possible to detect fatty acids
such as acetic acid at low concentrations using such gas sensors.
The present invention provides new gas sensors which employ new
semiconducting organic polymers With these polymers, high
sensitivity towards fatty acids (such as acetic acid) and ammonia
can be achieved.
The new materials have a bilayer structure with a baselayer of
polypyrrole deposited chemically using ferric chloride as an
oxidant. Different sensor types are manufactured by
electroechemically depositing different top layer polymers onto
this baselayer. The four types of sensors incorporated into the
device described above use the following monomer/electrolyte
combinations for the electrochemical deposition stage:
1. 3-Hexanoylpyrrole/tetraethylammonium p-toluenesulponate
2. 1-Octylpyrrole/tetrabutylammonium triflate
3. 3-Dodecylpyrrole/tetraethylammonium tetrafluoroborate
4. 1-Dodecylpyrrole/tetraethylanmmoniun tetrafluoroborate
The 3-substituted monomers can be synthesised following the method
of Ruhe et al (Makromol, Chem., Rapid Commun. 10 (1989) 103). The
1-substituted monomers can be synthesied following the method of
Santaniello et al (Synthesis, 1979, 617).
Further details of the polymerisation conditions and of the
preparation of polymer bilayers having a baselayer of polypyrrole
can be found in the Applicant's earlier International Publication
WO 96/00383.
In a typical procedure, the sample is equilibrated at 40.degree. C.
for ca. 3 minutes prior to sampling to allow a consistent
generation of the sample headspace. Nitrogen gas is humidified to
50% relative humidity and introduced into the sample vial directly
above the surface of the swab. The sample headspace is delivered to
the sensor array for 3 minutes at a flow rate of ca. 60
mlmin.sup.-1. The sensor array is allowed to recover for a few
minutes before a "wash" gas, preferably a high concentration acetic
acid wash, is passed over the sensor typically for 1 to 4
minutes.
Data Processing
An object of the invention is to produce a rapid screening system
for BV. The embodiment described below utilises the acetic acid,
ammonia and TMA sensitive sensor array described above and is
capable of making a decision based on the relative intensities of
acetic acid and/or ammonia and TMA present in the headspace.
Analysis of the sensor array has been greatly facilitated by a
novel data processing technique which is discussed below and which
is based on principal components analysis (PCA--see, for example, J
E Jackson, J Qual. Tech., 13(1) (1981)). It has been established
that if a principal components analysis of the intensity data from
the highly ortogonal sensors is carried out, the distribution of
points projected on a first principal components axis PC 1 is
correlated to acetic acid, and that the points projected on a
second principal components axis PC 2 are correlated to ammonia.
The distribution along either coordinate axis is also a function of
the concentration of the analyte in the headspace, and hence of the
concentration of marker chemicals produced by the microorganisms
present in the sample. Thus it is possible to utilise a
thresholding technique for deciding whether or not a sample is
positive or negative, based on user-defined clinical criteria. In
broad terms, the data processing comprises using calibration
samples to establish a PCA "calibration map", and then projecting
data obtained from enclosed samples onto this PCA calibration map
in order to establish if these data are indicative of
infection.
Data processing is described in more detail below, with reference
to various calibration and measurement process which are
performed.
1. Calibration
Calibration involves running calibration standards to generate a
reference map, verifying results and storing PCA loadings
information.
In one example the array is calibrated using defined standards
consisting of a "blank" sample, two acetic acid standards of
different concentrations, and ammonium hydroxide. Typically, the
calibration process is accompanied by repeatability checks on the
standards in order to verify tat there is sufficient discrimination
between standards and that repeatability is within acceptable
bounds. 1 ml sample volumes are used as above. As part of each
experimental run, subsets of standards are run, each sample cycle
lasting 20 minutes. In preferred, but non-limiting, embodiments,
the blank is water, one acetic acid standard comprises 900 ppm
acetic acid in 0.01M HCl, the other acetic acid standard is 5000
ppm acetic acid in 0.01M HCl, and the ammonium hydroxide standard
is 10 ppm in 0.01 M NaOH. The subset of standards can be used to
confirm that the performance of the system is still the same as
during its calibration and hence that the samples can be
processed.
1.a. Calibration Map
The calibration data is transformed using Principal Component
Analysis (PCA) to characterise the instrument sensor responses for
the calibrants run, thus defining a two-dimensional mapping space
on to which all subsequent samples analysed can be projected. PCA
decomposes the original calibrant data matrix into a set of scores
and loading vectors, in which scores contain information of how
samples relate to one another whilst the loadings show how
variables relate to one another. This process is depicted in FIG.
3, and can be written as:
where t denotes the scores, which are vectors of linear
combinations of sensor variables that describe the major trends in
the original data matrix X. The loadings, which are represented by
p, are a set of orthonormal eigenvectors representing a new set of
axes onto which the scores information is projected. In FIG. 3 T
denotes the transpose of a matrix. The result is a "calibration
map" which is shown in FIG. 4.
Projected scores (PC1 and PC2) can be calculated by multiplying the
analysis results with the reference loadings calculated in the
calibration step. This process is depicted in FIG. 5.
1.b. Setting Thresholds
In the following discussion, Blk is the water standard referred to
above, Std1 is the lower concentration acetic acid standard, Std2
is the higher concentration
Where A,B,C,D,E and F are constant terms the values of which are
determined by the parameters (r, s', c, d, M(.theta..sub.r),
N(.theta..sub.r)) such that:
The parameters M(.theta..sub.r) and N(.theta..sub.r) are defined
below and are dependent on .theta..sub.r.
The five parameters r, s', c, d, .theta..sub.r can be determined
from the set of calibration data. FIG. 6 shows the two dimensional
PCA reference map with the Blk 60, Std162, Std264 and Std366
calibration data projected thereon. FIG. 6 also depicts the
parameters r, s and .theta..sub.r and the resultant ellipse 68.
Each calibration sample is defined by a score (t1, t2) value along
each principal component axes PC1 and PC2, while (t1.sub.x,
t2.sub.x) denotes the mean score for a of n-repeats of a chemical
standard X. The length s' is the resolved distance (between the
mean scores for Std3 and the mean scores for Blk) orthogonal to
vector P determined by vector s which is at an angle .phi. to
vector r.
The five parameters r, s', c, d, .theta..sub.r are calculated as
follows:
Angle of Rotation .theta.acetic acid standard referred to above,
and Std3 is the ammonium hydroxide standard referred to above. Std1
and Std3 correspond to the threshold concentrations of acetic acid
and ammonium, respectively, above which infection is considered to
be present.
It is possible to set threshold levels which, if exceeded, are
taken to be indicative of infection. A relatively straightforward
way of doing this is to define a threshold PC1 value and a
threshold PC2 value. However, the present invention provides an
improved thresholding technique which generates an ellipse, and
utilises the boundaries of the generated ellipse as a threshold. An
advantage with the use of an ellipse for this purpose is that the
effects of sensor drift, which inevitably occurs over a period of
time, can be taken into account.
An ellipse is defined by the five parameters, r, s', c, d,
.theta..sub.r, where:
r is the radius of the major axis
s' is the radius of the minor axis
c and d relate to the coordinates of the centre of the ellipse
and
.theta..sub.r is the angle of rotation between the positive x axis,
and the axis radii established by Std1.
More specifically, an ellipse can be expressed algebraically in the
form:
From FIG. 6 the angle .theta. is defined as; ##EQU1##
which is calculated in radians.
Parameters c and d (a Function of the Blk Centre)
The mean score of tie Blk (t1.sub.Blk, t2.sub.Blk) standards
defines the centre of the ellipse The parameters c and d which are
functions of both t1.sub.Blk, and t2.sub.Blk are calculated as
follows: ##EQU2##
Axis Radii (r)
The length r is calculated as follows.
where r is the Euclidean distance separation between the mean
scores for std1 (t1.sub.Std1, t2.sub.Std1) and mean scores Blk
(t1.sub.Blk, t2.sub.Blk).
Axis Radii (s')
The general equation of an ellipse centred at (t1.sub.Blk,
t2.sub.Blk) with axis radii (r,s') rotated at an angle
.theta..sub.r from the positive x-axis can be written as follows:
##EQU3##
Where:
x=x-coordinate lying on the ellipse
y=y-coordinate lying on the ellipse
By re-arranging equation (4) the length of s' can be determined for
known values of c, d, r, .theta..sub.r and a point lying on the
boundary of the ellipse. The (x,y) coordinates in equation (4) are
defined by the mean score of Std3 (t1.sub.Std3, t2.sub.Std3) such
that: ##EQU4##
In general, a number of calibration measurements are made, and some
variation is the recorded responses is observed. This data scatter
is depicted in FIG. 6. It is possible to account for this variation
in the recorded responses for the calibration standards by
calculating a confidence band around the ellipse. To achieve this
the variability sd.sub.r, and sd.sub.s along the response vectors r
and s for the axis radii parameters are calculated.
Calculation of sd.sub.r and sd.sub.s along r and s for the axis
radii parameters.
A standard deviation value can be calculated for each standard
Std1, Std3 along each component axis. This can be resolved as shown
in FIG. 7 below to provide an estimate for the standard deviation
(sd) along the vector component r and s.
The resultant sd (.kappa..sub.r) along r(Std1) is:
and the resultant sd (.kappa..sub.s) along s (Std3) is:
Where sd.sub.t1 is SD along PC1 and sd.sub.t2 is SD along PC2.
The standard deviation along vector r (or s) is given by:
where: ##EQU5##
Having established sd.sub.r and sd.sub.s along the response vectors
r and s, the upper and lower lengths for r and s can be calculated
at the desired confidence level such that:
The confidence level (.alpha.) is calculated for two-tailed
t-statistic for v degrees of freedom, where .nu.=n-1 for n-repeats
of each standard calibrant.
Before the equations for the upper and lower confidence bands can
be established the coordinates for the lens Us and Ls are
determined.
Calculating the PC scores for a point lying along the vector s
whose length is Us and Ls from the ellipse centre.
Vector s can be written as:
Where
The length of s is given below:
The PC scores for a leng P along s can be calculated as
follows:
Where F is a multiplication factor and is defined as: ##EQU6##
Note that for P we can substitute the lengths Us and Ls in order to
determine the corresponding co-ordinates for U.sub.s' and L.sub.s',
respectively, from equation (11). U.sub.s' and L.sub.s' are
calculated as follows:
2. Classification
Once the parameters (Ur, c, d, .theta..sub.r, Us'(.alpha.)) and
(Lr, c, d, .theta..sub.r, Ls' (.alpha.)) have been determined, the
equations representing the upper and lower confidence intervals for
the ellipse can be defined. Real data are projected onto the PCA
calibration map and compared to the ellipse generated from the
calibration data.
The present inventors prefer to utilise the boundary of the ellipse
itself as the threshold for classifying a sample as being positive.
In other words, if a sample produces a PCA result which lies within
the ellipse, it is classified as negative, whereas if a sample
produces a PCA result which lies outside tie ellipse, it is
classified as positive for BV. However, the skilled reader will
appreciate that there are sundry variations in the way in which the
threshold might be set. For example, the lower confidence intervals
of the ellipse might be used as the threshold. ##EQU7##
system at regular intervals. This process is known as a system
verification process. Sensor responses can vary over a period of
time, and this response variation is often manifest as a rotation
of Std1, Std2, and Std3 responses about the origin of the PCA
calibration map. An advantage with the use of an ellipse to define
threshold values (as opposed to, for example; a linear threshold on
either PC axis) is that such rotation can be accommodated. However,
it is possible that sensor drift (or other factors) may cause so
large a shift in sensor response that the Std1 and Std3 responses
are unacceptably far removed from the ellipse when these responses
are transferred onto the PCA calibration map. In this instance, the
entire process can be repeated, and a new PCA calibration map
generated. A system verification process is now described.
A lower limit ellipse classification boundary (LECB) and an upper
limit ellipse classification boundary (UECB) are useful panmeters
in conducting system verification checks. The algebraic expression
describing the LECS and UECB can be determined by the following
sets of parameters, such that:
Note that the parameters c, d, and .theta..sub.r are already
calculated.
Calculating Lr and Ur
Lr and Ur are functions of the NSr parameter which is express in
standard deviation units from the mean of the Std1 standard along
vector r.
Calculating the PC coordinates for lengths Ls and Us along s.
Ls and Us are both functions of the NSs parameter which is
expressed in standard deviation units from the mean of the Std3
standard along vector s.
s can be written in vector notation as:
where .alpha.=t1.sub.Std3 -t1.sub.Blk and b=t2.sub.Std3 -t2.sub.Blk
(See equation 10)
The PC1 and PC2 coordinates for the length Ls and Us along vector s
is calculated as follows: ##EQU8##
Calculating Ls' and Us' ##EQU9##
where values for M and N are as calculated earlier.
The LECB and UECB is used during system verification. In one
embodiment, a number of Std1, Std3 and Blk standards are run. Std1
and Std3 standards should produce PCA results which lie within the
interval defined by the LECB and UJECB, ie, outside the LECB but
inside the UECB. The Blk standards should produce PCA results which
lie wit the LECB. An acceptable protocol is to run three Std1, tree
Std3 and three Blk, and to require that all of the Blk standards,
at least two of the Std1 and at least two of the Std3 standards
satisfy these criteria. If this is not the case, then the system
should be recalibrated. The system verification can be performed at
any suitable juncture, such as before a batch of samples is
analysed or perhaps on a daily basis.
Variants to the scheme described above would suggest themselves to
the skilled reader. For example, samples may be separated and
classified using Mahalanobis distance measure. Other forms of data
reduction than Principle Components Analysis might be used, such as
Sammon Mapping. In principle, more than two principal components
might be used to construct the calibration map. Such an approach
may not be of great advantage in the context of the technique for
detecting BV discussed above. However, the data processing
principles discussed above may be applicable to the analysis of gas
sensors in other application areas and the approach may even be
applicable beyond the field of gas sensors, perhaps to the analysis
of data from combinations of other kinds of sensor, or to
multiva-rate data analysis per se, In the instance where acetic
acid alone is being detected as a "marker" for BV, it may be
appropriate to utilise a single PCA axis, i.e., to consider a
single principal component. Other forms of data analysis, such as
neural networks or chemometric techniques, might be used. It will
be appreciated by the skilled reader that, although the calibration
map is a useful device for visually displaying the results of
calibration and subsequent real data, for the purposes of
calculating whether a sample is positive or negative it is not
essential to display such a map. Rather, the method of performing
data reduction to provide calibration loadings which are used to
transform the output of the detector so tat an assessment of
whether BV is present can be performed entirely in software,
without actually constructing a map per se.
Instead of detecting gaseous and volatile species in the gas phase,
it may be possible to perform measurements in the liquid phase.
Spectroscopic techniques such as infxa-red spectroscopic might be
used. Alternatively, continuous flow methods, electrochemical
techniques, or analysis by measurement of electrical impedance,
such as described in International Publication No. WO 98/96985,
might be contemplated.
Having described and illustrated the principles of the invention in
a preferred embodiment thereof, it should be apparent that the
invention can be modified in gement and detail without departing
from such principles. We claim all modifications d variation coming
within the spirit and scope of the following claims.
* * * * *